Give your AI assistant a safe path to Deniable.

What MCP does

The Model Context Protocol gives AI applications a standard way to discover tools and resources. Deniable exposes its catalogue through read-only tools so an assistant can find the right dataset before a metered retrieval is requested.

Connect from the AI client you already use.

The remote MCP pattern is designed for Claude, ChatGPT, Cursor, VS Code, Gemini, and other AI agents that support Streamable HTTP MCP connections. Each client may use a different settings screen or authentication flow.

Production endpoint: https://api.deniable.net/mcp. Store the Deniable API key in the client's secure authorization configuration — never in a prompt, repository, browser bundle, or URL.

Available tools

• list_datasets — search published catalogue metadata; never returns document content, storage paths, or raw files.
• retrieve_documents — retrieve pre-generated JSON documents and charge the applicable credits.
• check_usage — inspect the authenticated account's balance and delivery totals.

Credit and delivery rules

Standard items cost 1 credit. Long-form documents and ebooks cost 10 credits. The minimum credit top-up is €5, with no monthly subscription. A request stops before delivery when the balance is empty.

Deniable tracks delivered document IDs per API key and favours unseen items before beginning another cycle. A new retrieval is a new delivery and consumes credits again.

Security boundary

Catalogue search returns metadata only. It does not reveal the document body, preview rows, R2 object keys, filenames, or download URLs. Retrieval requires a valid API key, an API-enabled dataset, sufficient credits, and a successful rate-limit check.

Once an authorised client receives a document, it can technically copy or retain that response. No API protocol can prevent a customer from saving data they were allowed to retrieve. Authentication, least-privilege keys, metering, rate limits, delivery history, and clear licensing are the practical controls.

Get a step-by-step tutorial from your AI assistant

Copy the prompt below and paste it into Claude, ChatGPT, Cursor, VS Code, Gemini, or another AI assistant. Then paste the full Deniable MCP documentation after the prompt.

Act as an integration expert for the Deniable MCP Server. Using the Deniable MCP documentation I provide below, create a step-by-step tutorial for connecting it to my AI client or project. Write in plain language and include the exact configuration fields, authentication steps, example tool calls, credit rules, rate limits, and security precautions. Cover Claude, ChatGPT, Cursor, VS Code, Gemini, and a generic Streamable HTTP MCP client where the setup differs. Explain how to search catalogue metadata without exposing document content, how to retrieve documents only when needed, and how to handle empty credits, authentication failures, rate limits, and retries. If any information is missing from the documentation, ask me before making assumptions.

Here is the Deniable MCP documentation:
[PASTE THE FULL DENIABLE MCP DOCUMENTATION HERE]
`